Michigan Consumer Confidence and Share of the 2-Party Vote

[NHPolitico]

I always thought before Vorlon's post from Public Opinion Strategies that there was a clear relationship between The University of Michigan's Consumer Confidence Index and the outcome of a presidential re-election.  I liked it because I thought it might collect all sorts of issues and dump them into one value. I've studied it and I think it does. I'll try to quantify it here.

The first thing I noticed after running my numbers is that Carter's 1980 re-election didn't fit the model well-- until October. So, if this year is like 1980, then we'll have to wait until October to predict Bush's share of the 2-party vote. Otherwise, if the dynamics are like 1984 or 1992 or 1996, then we can predict the results now.

What I did was to just divide the Election Day share of the 2-party vote for the incumbent into the consumer confidence rating.  This gave me an index value between 1.576 (for Bush in October 1992) and 1.764 (or Clinton in October 1992).  I grouped the monthly values for 1984, 1992, and 1996 together (to compare apples to apples) and averaged them to try to smooth out the index value into a single monthly index value-- so if you're using consumer confidence numbers from month X, use the index value from that month.  This wouldn't work well if the values varied by much within the months based on who the president was. Fortunately, that wasn't the case. The lowest monthly average index value was 1.656 for October and the highest was 1.687 in September-- a pretty narrow range. The average of all the monthly averages is 1.6729.  

So, I took that 1.6729 and divided it into the monthly consumer confidence rating for all three incumbents for every month to get a predicted share of the 2-party vote. The range of difference between the actual and predicted share of the two-party vote using the same 1.6729 value to predict every president for every month is 2.98 points too optimistic for Clinton in October 1996 and 2.68 points too pessimistic for Bush in October 1992.  That's better than most public opinion polls.  The best month to predict the share of the 2-party vote is May when the average absolute value difference between actual and predicted share is 0.887 points.

Now, as far as Carter, in September and October the race seemed to settle into the same pattern as 1984, 1992, and 1996 with index values of 1.649 and 1.678 respectively.  Before then, it really varies between 1.157 in May and 1.506 in August. Carter's average index value is 1.392. If you use that to predict Carter's share of the vote in every month using that month's confidence index number, it's accurate in July, but otherwise it's garbage-- from 9.2 points too optimistic in October to 7.6 too pessimistic in May.  Carter's average index value for September and October is 1.6633 and using that value along with the consumer confidence values for September and October separately pegs it right: +/- 0.4 points (damn good).  So, basicly, if this predictor fails to predict Bush's performance, it will likely mean that the dynamics are more like 1980 than the last three re-election campaigns.

So, the prediction of Bush's 2004 share of the 2-party vote as of April's preliminary Michigan Consumer Confidence Index (if the dynamics are like the last three re-election races) is: 55.7% vs. 49.7% in 2000.  Who says we're a 50:50 nation?

[NHPolitico]

And this will probably help raise the consumer confidence (and his share of the 2-party vote) higher:

"The Commerce Department reported today U.S. orders for durable goods rose 3.4 percent in March, almost five times economists' forecast, led by bookings for autos, metals and business equipment."